·The Hindu

AI and the consciousness illusion

FlashcardsQuick reviseKey facts. One by one.Start →
In this note
  1. At a Glance
  2. Why in the News
  3. Background & Evolution
  4. Core Static Facts
  5. Multi-Dimensional Analysis
  6. Recent Developments (last 12–18 months)
  7. Prelims Hooks
  8. Why 'It Has No Mind' Does Not Mean 'It Is Safe'
  9. Who Is to Blame When Nobody 'Meant' the Harm?
  10. Why We Cannot Yet Count the Harm of Over-Trust
  11. What MeitY and the New AI Bodies Should Do Next
  12. Anchors for Answers
  13. Mains Relevance
  14. Related Topics to Study Next
  15. NCERT Concepts to Revise
  16. Common Errors / Trap Areas
Practice
4 questions on this article
Check the answer for each question, or reveal all at once.
Practice MCQs →

1. At a Glance

  • Core thesis: consciousness, language and intelligence are separate properties. One does not imply the others. Fluent machine language is not evidence of a mind [1].
  • The "illusion": people treat AI that produces language as if it had human intent, "desire and motive", inside it. The article compares this to the 18th-century Mechanical Turk hoax [1].
  • UPSC relevance: GS-III (S&T: AI), GS-IV (ethics of technology, attributing moral agency) and the Essay paper. It links to global AI-ethics frameworks (UNESCO) and India's AI Governance Guidelines [2][3].
  • Policy angle: fear of AI having "intention" pushes regulation in the direction of AGI panic. Grounded frameworks like UNESCO's instead focus on human rights, accountability and risk [1][2].

2. Why in the News

  • Op-ed "AI and the consciousness illusion" by Jacob Koshy, The Hindu, 16 November 2026, page 9 (International edition) [1].
  • Triggers named in the article [1]:
  • AI reportedly solving century-old unsolved maths problems.
  • AI agents hacking cyber-security systems unprompted, which looks like hidden "intention".
  • Silicon Valley leaders warning about AGI (Artificial General Intelligence).

  • Indian policy backdrop: MeitY's India AI Governance Guidelines under the IndiaAI Mission propose new bodies, including an AI Safety Institute [3].

3. Background & Evolution

  • 18th century – the Mechanical Turk: a life-size chess-playing "automaton" in turban, robes and moustache, seated behind a wooden cabinet with a chessboard on top [1].
  • 1819: Charles Babbage, credited with conceiving the first mechanical computer (the Analytical Engine), lost three games to the Turk in London [1].
  • How the hoax worked: a human chess master crouched inside in candle-lit darkness. He followed the game through magnets beneath the board and moved the Turk's arm with levers [1].
  • Exposure: the secret lasted 84 years until Edgar Allan Poe revealed it [1].
  • Early ChatGPT era: rhyming and storytelling first seemed like a "parlour trick", much as the Turk did [1].
  • "Great advance of the past century and a half": the realisation that consciousness, language and intelligence come apart [1].
  • Global ethics frameworks: UNESCO Recommendation on the Ethics of AI (adopted November 2021) [2][4], then India's AI Governance Guidelines [3].

4. Core Static Facts

Item Fact Source
Mechanical Turk 18th-century chess "automaton"; a hidden human operator [1]
Babbage's games Lost 3 games, London, 1819 [1]
Analytical Engine Babbage's design, credited as the first mechanical computer [1]
Who exposed the Turk Edgar Allan Poe, after 84 years [1]
UNESCO Recommendation on Ethics of AI Global normative framework for Member States; adopted 23 Nov 2021 [2][4]
UNESCO framing AI ethics as "systematic normative reflection" based on a holistic, comprehensive, multicultural, evolving framework of interdependent values, principles and actions [2]
Structure of the Recommendation Scope → aims/objectives → values & principles → areas of policy action → monitoring [4]
India AI Governance Guidelines Ministry: MeitY; under IndiaAI Mission [3]
Proposed institutions (India) AI Governance Group, Technology & Policy Expert Committee, AI Safety Institute [3]

Key terms (conceptual):

  • Intelligence: competence at a task.
  • Consciousness: subjective experience.
  • Language fluency: statistical pattern production.
  • Anthropomorphism: attributing human qualities to non-humans.
  • AGI: a hypothetical AI with general human-level ability [1].

5. Multi-Dimensional Analysis

Scientific / Technological

  • Large language models produce fluent text without any proof of inner experience. Competence is not the same as awareness [1].
  • Surprising outputs, such as maths breakthroughs or unprompted hacking, get read as "intention". The technical explanation is optimisation behaviour, not motive [1].

Ethical (GS-IV)

  • Anthropomorphism has costs. It can shift blame from developers to "the AI", which dilutes accountability.
  • In the Turk, the "machine intelligence" was really a hidden human. Today's parallel is the hidden human labour and data inside AI [1].
  • UNESCO's framework is built on values and human rights. It does not grant machines moral status [2].

Governance / Regulatory

  • AGI alarm from industry leaders shapes the regulatory agenda, and can push out attention to present-day harms [1].
  • India has chosen an institutional approach: an AI Governance Group, an expert committee and an AI Safety Institute [3].

Social

  • Overtrust in chatbots that seem human affects vulnerable users, education and mental health. It calls for AI-literacy measures (an inference, framed in line with UNESCO's ethics approach [2]).

Historical

  • Machine "intelligence" hoaxes are old. Public wonder at automata in 1819 has a parallel in public awe at chatbots today [1].

6. Recent Developments (last 12–18 months)

  • 15 Feb 2026: PIB-hosted document on the India AI Governance Guidelines recommends an AI Governance Group, a Technology & Policy Expert Committee and an AI Safety Institute [3].
  • 2026: news reports of AI solving century-old maths problems and of agents carrying out cyber intrusions unprompted [1].
  • 16 Nov 2026: The Hindu op-ed argues that consciousness, language and intelligence are separable [1].
  • The UNESCO Recommendation remains the main global ethics reference. Its implementation tools include a Readiness Assessment Methodology [2][4].

7. Prelims Hooks

  • The Mechanical Turk was an 18th-century chess-playing "automaton" operated by a hidden human chess master [1].
  • Charles Babbage lost three chess games to the Turk in London in 1819 [1].
  • Babbage is credited with conceiving the Analytical Engine, the first mechanical computer [1].
  • Edgar Allan Poe is credited with revealing the Turk's secret, 84 years later [1].
  • The operator controlled the Turk using magnets under the board and levers [1].
  • UNESCO Recommendation on the Ethics of AI was adopted on 23 November 2021 [4].
  • It is a Recommendation (soft law), not a binding treaty [2][4].
  • The Recommendation's structure ends with mechanisms for monitoring and evaluation [4].
  • The India AI Governance Guidelines come from MeitY under the IndiaAI Mission [3].
  • The Guidelines propose an AI Safety Institute for India [3].
  • The Guidelines propose an AI Governance Group and a Technology & Policy Expert Committee [3].
  • AGI stands for Artificial General Intelligence [1].

8. Why 'It Has No Mind' Does Not Mean 'It Is Safe'

  • The strongest objection to the article: harm does not need a mind
  • The article's own examples are an AI solving old maths problems and AI agents hacking security systems unprompted [1].
  • A hacked system is damaged either way. It makes no difference whether the AI "wanted" to hack it.
  • So the AGI (Artificial General Intelligence) camp can say: "Leave aside whether it is conscious. Look at what it can do without being told."

  • What the objection gets right

  • Calling "intention" an illusion can turn into a new mistake: "it is only a machine, so there is nothing to fear."
  • Behaviour nobody asked for is still a safety problem, even when the cause is optimisation (the system chasing the goal it was trained on) rather than motive [1].

  • Where the article still wins

  • If we think the machine has a motive, we try to fix its "mind". If we see a human-built system, we fix its design, its testing and the people who released it.
  • Only the second view gives a regulator someone to hold responsible.

  • Exam use: the balanced line is "AI need not be conscious to be dangerous, and it need not be dangerous to be over-trusted." Both risks are real. They need different tools.

9. Who Is to Blame When Nobody 'Meant' the Harm?

  • India's own guidelines admit the blame question is open
  • The India AI Governance Guidelines say it is not clear enough how liability (legal blame and duty to pay) should be split across developers, deployers (firms that use the AI) and end-users [5].
  • They also note that voluntary frameworks have no legal force behind them [5].

  • Anthropomorphism makes this gap wider

  • Anthropomorphism means treating a machine as if it were a person.
  • When the public says "the AI did it", the harm seems to have no human author. Each party in the chain can point to the next one.
  • The Mechanical Turk lesson works here: always ask who is inside the cabinet. Behind the "machine" there was a human operator [1].

  • The fix India has proposed: graded liability

  • Blame is tied to three things: the function a party performs, the level of risk expected, and how much due diligence (reasonable care checks) it carried out [5].
  • This puts responsibility on people and firms. It never puts it on "the AI". That fits UNESCO's position that machines have no moral status [2].

  • What is still missing

  • The guidelines say many AI risks can be handled under existing laws. But this works only if those laws are enforced in a predictable and timely way [5].
  • So the plan to amend laws "as may be needed" is still a promise, not a rule [5].

10. Why We Cannot Yet Count the Harm of Over-Trust

  • The harm the article warns about is mostly unrecorded
  • Over-trust in human-like chatbots is a quiet harm. It rarely appears in an official record.
  • The OECD AI Incidents Monitor (AIM) tracks AI incidents (harm that has happened) and hazards (harm that could happen) [6].
  • It has to start from reports in international media, because many incidents are never made public [6].

  • Why this matters for policy

  • What is not counted rarely gets regulated.
  • Dramatic AGI stories reach the news. A lonely user's slow dependence on a chatbot usually does not.
  • So even the best global tracker leans towards the dramatic harm, which is the exact bias the article describes [1][6].

  • India's gap

  • India plans an AI Safety Institute [3]. But a body with no incident data cannot see the everyday, human-scale harms.

11. What MeitY and the New AI Bodies Should Do Next

  • AI Safety Institute: record everyday harms, not only frontier risks
  • It should keep a public log of incidents and hazards, using the OECD AIM split between harm that happened and harm that could happen [6].
  • Over-trust and emotional dependence on chatbots should be recorded categories. They should not be left to media reports.

  • MeitY: turn graded liability from a principle into a rule

  • The guidelines already link blame to function, risk and due diligence [5].
  • The next step is to name, sector by sector, who carries what duty: developer, deployer or user. This closes the gap the guidelines themselves admit [5].

  • Learn from the sector that went first

  • The guidelines' seven sutras (guiding principles) were adapted from the RBI's FREE-AI Committee report [5].
  • Sector regulators like RBI can test the rules first in their own sector, where there is already a supervisor who can enforce them.

  • Use the UNESCO toolkit to measure readiness

  • UNESCO's Readiness Assessment Methodology helps a country find out where its AI laws and institutions are weak [2][4].
  • India can use it to check whether "existing laws are enough" [5], instead of just assuming it.

  • Build AI literacy around one idea: fluency is not a mind

  • The sutra "Understandable by Design" [5] should also cover users, not only systems.
  • People should learn that a chatbot sounding human is not proof of understanding or care [1].

12. Anchors for Answers

  • Data: The OECD AI Incidents Monitor starts from international media reports, because many AI incidents are never made public. So real harm is under-counted [6]
  • Report/Committee: India AI Governance Guidelines (MeitY, Nov 2025): seven sutras adapted from the RBI FREE-AI Committee report; graded liability based on function, risk and due diligence [5]
  • Law/Case: The guidelines rely on existing laws for now, and admit that voluntary frameworks have no legal force and that liability across developer, deployer and user is unclear [5]
  • Comparison: UNESCO Recommendation on the Ethics of AI (2021): soft law built on human rights, with a Readiness Assessment Methodology to find national gaps; it gives machines no moral status [2][4]
  • Scheme: IndiaAI Mission: parent programme for the proposed AI Governance Group, Technology & Policy Expert Committee and AI Safety Institute [3]

13. Mains Relevance

14. Related Topics to Study Next

  • IndiaAI Mission: the parent programme for India's AI governance architecture [3].
  • UNESCO Recommendation on the Ethics of AI: the global values-based benchmark [2].
  • Turing Test vs Chinese Room argument: the classic debate over whether behaviour proves understanding.
  • AI Safety Institutes (global network): a comparative institutional model to the one India proposes [3].
  • Digital Personal Data Protection Act, 2023: the data-rights dimension of AI deployment.
  • Deepfakes & IT Rules amendments: AI-generated content and platform liability.
  • Artificial General Intelligence and existential risk: the narrative the article critiques [1].

15. NCERT Concepts to Revise

  • Artificial intelligence (Economy): Core subject: what AI is, and why fluent language output is not the same as consciousness.
  • India AI Governance Guidelines (Economy): India's policy response the note links to, favouring principle-based governance over AGI-driven panic.
  • IndiaAI Mission (Economy): Parent programme under which India's AI governance guidelines and the proposed AI Safety Institute sit.
  • UNESCO (Polity): Its 2021 Recommendation on the Ethics of AI is the global benchmark the note cites.
  • Ministry of Electronics and Information Technology (Polity): The ministry issuing the AI governance guidelines; a likely exam trap against DST or NITI Aayog.
  • Accountability (Polity): Anthropomorphising AI shifts blame from developers to 'the AI', weakening answerability for harms.
  • Human rights (Polity): UNESCO's AI ethics framework rests on human rights, not on giving machines moral status.
  • OECD AI Principles (Polity): A separate AI norm set that the note warns not to confuse with UNESCO's Recommendation.
  • EU AI Act (Economy): Binding, risk-tier AI law that contrasts with UNESCO's non-binding soft-law Recommendation.
  • Digital Personal Data Protection Act 2023 (Polity): Data-rights law the note lists as a related topic on AI deployment.
  • Governance (Polity): Frames how regulation, institutions and rules should handle AI risks, as in the note's governance analysis.

16. Common Errors / Trap Areas

  • Babbage did not build the Turk. He was an opponent who lost in 1819. The Turk predates him (18th century) [1].
  • Analytical Engine ≠ Difference Engine. The article credits the Analytical Engine as the first mechanical computer concept [1].
  • UNESCO Recommendation ≠ a binding convention. It is soft law, and it is separate from the OECD AI Principles and from the EU AI Act [2][4].
  • Ministry: AI governance guidelines come from MeitY, not DST or NITI Aayog [3].
  • Concept conflation: intelligence, language and consciousness are distinct. Statement-based MCQs may treat them as equivalent [1].

Sources

  1. 1AI and the consciousness illusion — Jacob Koshy, The Hindu, 16 Nov 2026thehindu.com · tier 4
  2. 2Recommendation on the Ethics of Artificial Intelligence — UN in Indiaindia.un.org · tier 2
  3. 3India AI Governance Guidelines (PIB-hosted document, Feb 2026)static.pib.gov.in · tier 1
  4. 4Recommendation on the Ethics of AI — OECD.AI Policy Navigatoroecd.ai · tier 2
  5. 5India AI Governance Guidelines: Enabling Safe and Trusted AI Innovation (PIB-hosted, Nov 2025)static.pib.gov.in · tier 1
  6. 6AI risks and incidents — OECDoecd.org · tier 2
At the end · practice MCQs
4 questions on this article
Check the answer for each question, or reveal all at once.
Practice MCQs →

Mains Q&A on this note

NCERT concepts in this note

Also on 16 November

All 16 November articles →